RECRUITING HIGHLY QUALIFIED TEACHERS: DO DISTRICT RECRUITMENT PRACTICES MATTER?

Transcription

1 RECRUITING HIGHLY QUALIFIED TEACHERS: DO DISTRICT RECRUITMENT PRACTICES MATTER? Dana Balter and William Duncombe* Education Finance and Accountability Program Center for Policy Research, Maxwell School of Syracuse University 426 Eggers Hall Syracuse, NY (315) FORTHCOMING IN PUBLIC FINANCE REVIEW *Corresponding author.

2 Recruiting Highly Qualified Teachers: Do District Recruitment Practices Matter? Abstract This paper presents results from a survey on teacher recruitment practices used in New York State school districts and analyzes whether the level of use of recruitment practices is related to teacher qualifications. We find that most districts employ a wide variety of practices and that the number of recruitment practices used by districts goes up with district size. To examine the effectiveness of recruitment practices, we estimate a model relating a composite measure of teacher qualifications to the level of use of recruitment practices and labor supply and demand factors. The recruitment regressor is treated as an endogenous variable with factors related to district use of these practices as instruments. While we cannot identify which individual practices are important, we find consistent evidence that districts using only a limited set of recruitment practices have hired less qualified teachers in New York. JEL Classification: H833, I22, J45

3 Recruiting Highly Qualified Teachers: Do District Recruitment Practices Matter? 1. Introduction Decades of research on the determinants of student achievement make it clear that high quality teachers matter to student success (Ferguson 1998; Goldhaber 2002; Hanushek, Kain, and Rivkin 2002). With states under pressure to raise teacher quality to comply with the No Child Left Behind Act (NCLB) and with a significant increase in demand for teachers projected for the next decade (Hussar 1999), many school districts face teacher recruitment challenges. The growing literature examining the determinants of teacher mobility and retention has focused on a limited set of factors within district control (e.g., wages, class sizes), and outside district control (e.g., demographics of student body, district location) that could affect mobility. Very little research exists, however, on what types of practices districts use to increase the number of applications for open teaching positions and on whether these recruitment practices are effective. Can school districts, through the strategic use of advertising, partnerships with colleges, and recruitment incentives, increase teacher quality? Using the results of a 2004 survey of school district superintendents in New York State on teacher hiring practices, the objectives of this paper are to: 1) describe the actual recruitment practices of New York school districts, and 2) examine the effect of teacher recruitment practices on the qualifications of teachers hired by New York school districts. We find that districts in New York use a range of recruitment practices, the number and types of practices vary with the enrollment size of the district, and districts employing only a limited set of recruitment practices have hired less qualified teachers, on average. While our results on effectiveness should be viewed as exploratory, they provide some of the first evidence that widespread use of teacher recruitment practices can improve teacher quality. 1

4 2. Teacher Recruitment Practices Research examining the determinants of teacher mobility decisions has focused primarily on two set of factors affecting teacher labor supply: relative salaries and working conditions. A large body of empirical research confirms that compensation can significantly affect teacher decisions about leaving and moving (Baugh and Stone 1982; Murnane and Olsen 1990; Stinebrickner 1998; Imazeki 2005; Dolton and van der Klaauw 1999). However, several recent studies suggest that teacher mobility decisions may be more strongly influenced by working conditions than by salaries, particularly the characteristics of the students (Hanushek et al. 2001; Scafidi, Sjoquist, and Stinebrickner forthcoming; Falch and Strom 2005). Furthermore, the recent finding by Boyd et al. (2004) that most public school teachers take their first public school teaching job very close to their hometowns or where they attended college (p. 117) suggests that teachers may have strong preferences for a familiar teaching environment Research on Teacher Recruitment Practices In contrast to the research on salaries and working conditions, very little attention has been paid to the practices districts actually use to recruit job applicants for teaching positions and whether these practices are effective in improving teacher quality. Teacher recruitment practices could increase the quality of teachers hired by the district for several reasons. First, if teacher job search, particularly beyond the local area, is limited by imperfect information about the school district, then districts may be able to increase applications through advertising in newspapers, journals, and other types of media (e.g., internet). Second, if applications are limited by the high entry costs into teaching created by state certification requirements and district policies, districts can reduce those barriers by encouraging applications from alternatively certified teachers and by 2

5 providing assistance for paraprofessionals to become teachers (Glazerman, Mayer, and Decker 2006; Boyd et al. 2006). Third, if uniform salary schedules reduce a district d flexibility in attracting teacher applications, particularly in hard-to-staff fields and schools, then incentives in the form of additional compensation or benefits might expand the applicant pool. The little evidence that does exist on actual recruitment practices in public education suggests that many districts engage in a fairly limited search for candidates. Based on a detailed survey in 1997 of superintendents, school board presidents, and teacher union presidents in Pennsylvania, Strauss et al. (1998; 2000) find that 75% of districts only advertise in Pennsylvania and 17% only advertise locally. The internet was used at least sometimes by 29% of respondents, and 30% of districts had partnerships with colleges for teacher training and placement. Strauss and colleagues (2000) conclude that districts in Pennsylvania focus their hiring process on substitute or part-time teachers and graduates of local colleges, instead of actively seek[ing] new teacher applications through vigorous advertising and recruiting (p. 405). Concerns over the quality of the public teacher hiring process have also been raised in several comparisons of teacher recruitment policies in public and private schools (Ballou 1996; Ballou and Podgursky 1998) Teacher Hiring Survey To shed some light on actual district hiring practices, we surveyed New York school superintendents in the spring of The survey covered three broad topics: 1) teacher recruitment, 2) teacher screening and selection, and 3) interest in training and support. For this paper we will concentrate on the results from the first section of the survey. In implementing the survey, we followed closely the recommendations of Dillman (2000) to maximize the response 3

6 rate. 1 The resulting response rate was over 71%. The survey sample did not include the New York City School District or districts primarily serving special populations. To examine how representative the sample is of all school districts in New York, we compared characteristics of responding districts to non-respondents on a range of demographic, fiscal, teacher, and region variables. Districts completing the survey have similar enrollment size and student socio-economic composition as non-respondents. The one exception is the share of limited English proficient (LEP) students, which was higher in districts not in the survey. Regarding district finances, districts in the survey have 6% lower spending overall and 3% lower operating spending. None of the other differences in fiscal variables are statistically significant from zero. The teacher work force in responding districts is very similar to non-respondents, except that non-respondents have slightly less experienced teachers and a slightly lower share of tenured and permanently certified teachers. Respondents and non-respondents tend to be distributed evenly between urban, suburban, and rural districts; however, respondents are less likely to be located in the New York City metropolitan area (so called downstate New York). 3. Use of Teacher Recruitment Practices in New York The high response rate and representative nature of the sample provides us the opportunity to examine in depth the teacher hiring practices of New York school districts. In this section we present simple bivariate relationships between use of specific recruitment practices and district characteristics including enrollment size and the fiscal health of a district. 2 The number of potential applicants aware of teacher openings in the district can be increased by using mass media, the internet, job fairs, and contact with colleges. Beyond broadening the pool of potential applicants, recruitment practices can help increase the number of applicants by providing financial 4

7 incentives and reducing barriers to entry into teaching for non-certified teachers and paraprofessionals. The survey results are summarized by the major categories of recruitment practices Advertising A simple and relatively inexpensive recruitment strategy, at least in terms of staff time, is to put advertisements in newspapers, trade publications, or on radio or television. The least expensive option would probably be advertising in local newspapers and possibly local access cable or radio stations. We asked districts to identify the media outlets they use and, in the case of newspapers, whether they are local (within 50 miles), in other areas of the state, or based out of state. 3 The typical district begins advertising in March or April for the following school year and makes an offer in June (Table 1); an early advertising date is associated with making earlier offers to prospective teachers. 4 High need urban districts begin advertising around the same time as average need and low need districts but make their offers one-half to one month later on average. 5 The principal media outlet for teacher job advertisements is local newspapers, with 80% of districts placing most or all of their advertisements for open teaching positions in local newspapers (Table 2). A much smaller share of districts (25%) place the majority of their advertisements in non-local newspapers within New York State. Less than 5% of districts make education trade publications, out-of-state newspapers, or radio and television their principal advertising media for teaching positions. Small districts are more likely to use local newspapers and less likely to use other New York newspapers than districts with high enrollment. A lower share of high need urban districts use newspaper advertising compared to other districts, but they are only slightly more likely to use advertisements in education trade publications. Unexpectedly, 5

8 high need rural districts are heavier users of radio and TV advertising than other districts, but even among rural districts use of radio and TV advertising is very limited. <Table 1 about here> 3.2. Recruiting from Colleges A potentially effective approach to recruiting new teachers is to work with the colleges producing them. The contact can be fairly passive, as in asking colleges to post job notices on bulletin boards or in placement newsletters, or it can involve more direct contact by visiting the campus or talking to faculty about job candidates. School districts and colleges can establish even stronger partnerships through student teaching arrangements and the interaction of college faculty and district personnel on curricular and pedagogical issues. We asked superintendents to categorize their district s relationships with local colleges (colleges within 50 miles), other New York State colleges, and colleges in other states. Districts work primarily with local colleges; the most common college recruitment strategies include posting job notices at the colleges, and more active strategies, such as supervising student teachers and contacting college faculty (Table 2). 6 A majority of districts also place job notices in placement newsletters and visit local campuses to actively recruit candidates. <Table 2 about here> With regard to non-local colleges (farther than 50 miles from the district), the only strategy used by a majority of districts is to post job notices at the college. Less than 40% of districts use active recruitment strategies with non-local colleges. Larger districts are more likely to recruit in both local and non-local colleges than smaller districts, particularly visiting campuses and supervising student teachers. High need urban districts are more likely to visit local colleges and 6

9 supervise student teachers from non-local colleges but otherwise are not any more active in working with colleges than other types of districts. As part of the survey, we asked superintendents to identify the five colleges with which you conduct the greatest number of these activities. Over 40% of recently hired teachers (hired in the last three years) in responding districts earned their bachelor degree at these colleges, and 55% earned their masters degrees. If local college is defined as a college in the same county or neighboring counties, close to 50% of recently hired teachers in New York received their bachelors in local colleges, and over 70% received their masters Use of the Internet The emergence of the internet provides new opportunities to expand teacher recruitment outside the local area at relatively low cost compared to traditional advertising. For the cost of posting job notices on teacher recruitment websites, assuming such sites are available, a district can potentially have access to a national market of teachers. Districts can post job notices on their own websites and can provide recruiting brochures online for candidates to download. Prospective teachers can communicate with the districts by and submit their applications online. Over 70% of districts use the internet, most commonly to post job notices on school district websites (Table 3). 8 Approximately 40% of districts also post openings on other teacher recruitment websites or allow teachers to submit their applications online. A much smaller share of districts actively use the internet to search for job candidates. High need urban districts are somewhat more likely than average need districts to use the internet, particularly to post job openings on district and teacher recruitment websites and to search for candidates on the web. While small districts are less likely to use the internet for recruiting, one exception is small districts that contract with a regional education organization in New York (Board of Cooperative 7

10 Educational Services, or BOCES) for access to online application systems or online placement of vacancy notices. 9 Small districts using BOCES online recruitment services are much more likely to use the internet for recruitment than other small districts, and their usage rates are similar to those of larger districts. <Table 3 about here> 3.4. Teacher Recruitment Incentives A range of teacher recruitment incentives has been discussed in the recruitment literature including signing bonuses, subsidized tuition, and assistance purchasing a home. 10 Districts can also offer extra compensation for teaching in hard-to-staff fields and schools, National Board Certification (NBC), or performing extra curricular activities and administrative duties as an inducement. Superintendents may also be able to increase base salaries by crediting teachers for experience in other districts or in non-teaching occupations. Almost three-quarters of superintendents responding to the survey said they used some type of recruitment incentive (Table 4), with the average district using between 1 and 2 incentives. 11 Only two incentives are used by over half the school districts extra compensation for supervising extracurricular activities and crediting teachers for experience outside the district. Two-thirds of districts use at least one of these traditional incentives. Approximately 16% of districts offer subsidized tuition at local colleges, additional compensation for NBC, or credit teachers for work experience in non-teaching occupations. The only other incentive used by at least 25 school districts is additional compensation in hard-to-staff fields. Slightly over 40% of districts use at least one of these non-traditional incentives. <Table 4 about here> 8

11 The use of incentives tends to go up with district size, particularly additional compensation for National Board Certification, and flexibility in crediting teaching experience in other districts. High need urban districts are more likely than other types of districts to use incentives, particularly signing bonuses, additional compensation for NBC, subsidized college tuition, and additional compensation for hard-to-staff fields or for extracurricular or administrative functions. It is interesting to note that these types of incentives have been recommended by several scholars as tools to attract high-quality teachers to high need schools, particularly in urban areas (e.g. Loeb 2000; Odden and Kelley 2000; Kearney 2000) Strategies to Increase Supply The findings of Boyd et al. (2004) that new teachers tend to teach close to their hometown, or where they went to college suggests that expanding the pool of teachers within the local area might be a more effective strategy to increase job applicants than broadening the job search beyond the local area. Strategies to increase the local supply of teachers might include recruiting substitute teachers, retired teachers, former teachers, and alternatively certified teachers or providing assistance for paraprofessionals to become certified teachers. Districts use two supply strategies on average, with recruitment of substitute teachers the most common strategy (Table 5). 12 Over 40%t of districts recruit teachers certified through alternative routes and 28% recruit retired teachers or provide assistance to paraprofessionals to become teachers. Only 7% of districts recruit former teachers. Use of supply strategies is not strongly related to district size, but larger districts are more likely to recruit substitutes and to assist paraprofessionals. High need urban districts are no more likely to use supply strategies than average need districts, except to recruit paraprofessionals. <Table 5 about here> 9

12 3.6. Use of Multiple Practices Most districts use a variety of practices, but only a small set of practices are used by the vast majority of districts. What is less clear is the portfolio of recruitment practices that districts employ, and whether practices of different types are treated as substitutes or complements. If a district posts job notices on the internet, for example, is it more or less likely than the average district to use advertising in newspapers or other media outlets as well? Table 6 reports for a district using a particular practice how likely they are to also use other practices compared to the average district. For example, if a district uses incentives in recruiting, they are 7.7% more likely than the average district to use the internet for recruiting as well. The predominance of positive percentages suggests that districts view most practices as complements when developing their recruiting plan. This is particularly true for less traditional practices, such as non-traditional recruiting incentives, other uses of the internet besides posting job notices on school websites, recruitment of alternatively certified teachers, and working with non-local colleges. Using one of these non-traditional practices is associated with 15% higher use of all other practices on average, and 22% higher use of other non-traditional practices. <Table 6 about here> 4. Relationship between Use of Recruitment Practices and Teacher Qualifications In this section, we propose an empirical model of teacher qualifications and present estimates from the model for New York school districts. As far as we are aware, this is one of the first attempts to examine the relationship between recruitment practices and the qualifications of recently hired teachers. Given that this analysis is based on cross-sectional data, and that proxy measures are used for teacher quality, the results should be viewed as exploratory. Despite these 10

13 limitations, the results provide evidence on the role that recruitment practices may play in improving teacher qualifications. 4.1 Measuring Teacher Qualifications The importance of teacher quality is frequently acknowledged in the education policy literature; however, measures of quality are often poorly defined and inconsistent (Goldhaber and Anthony forthcoming). Teacher quality would preferably be measured by student performance gains (Stone 1999; Sanders, Saxton, and Horn 1997), but the lack of micro-student data for New York requires that we take a more indirect approach by measuring teacher qualifications that may be related to teacher quality. While the extensive education production function literature has found inconsistent evidence about the relationship between teacher education, and experience and student performance (Hanushek 1986; Greenwald, Laine, and Hedges 1994), there is stronger evidence that certification in the subject taught is related to student performance gains that subject (Goldhaber and Brewer, 2000). There is also fairly consistent evidence that measures of teacher academic proficiency, such as performance on college entrance exams or teacher certification exams, are related to student performance gains (Ehrenberg and Brewer 1994; Ferguson and Ladd 1996). However, recent research by Goldhaber (2005) suggests that teacher certification exams may only explain a small percentage of student performance gains. To estimate teacher qualifications we use factor analysis with measures of teacher certification test score performance, certification status in teaching assignments, and college ranking to construct a composite measure of teacher qualifications (Loeb 2000). The data used in the factor analysis comes from the teacher certification database and the basic education data system maintained by the New York State Education Department (SED). 13 To match the timeframe in the survey we used information on recently hired teachers, which are defined as 11

14 teachers who began working in the district in the through school years. 14 Table 7 presents the factor analysis results, with the scoring coefficients (reported in the second column) used to construct the composite measure of teacher qualifications. 15 We estimated a Cronbach alpha to examine the reliability of the teacher qualification measure, and found that the reliability of this measure (Cronbach alpha=0.56) is below the standard (0.7) often used to identify reliable measures. <Table 7 about here> 4.2. Teacher Quality Model Observed teacher qualifications in a district reflect both district demand and teacher supply decisions. The cross-sectional nature of our data makes it difficult to estimate structural models of supply and demand for the teacher labor market. The approach we take instead is to estimate a reduced-form model of teacher qualifications, which includes both supply and demand factors. Teacher labor supply can be modeled as a function of four factors: teacher salaries (W) relative to comparable private sector salaries (P L ), working conditions (C), amenities (A), and personnel policies, such as recruitment practices (R) that may affect teacher supply (Hanushek, Kain, and Rivkin 2001). Most studies of teacher retention have focused on the first two categories salaries and working conditions and some hedonic models of teacher salaries have also considered amenities, such as urban location (Chambers 1995). District demand for teacher quality is a function of factors in an education cost model, including the level of student outcomes (S), factor price for teachers and capital (P K ), physical characteristics of the district (N) that could affect costs of production (enrollment and pupil density), and student characteristics (Z) affecting the translation of resources into student outcomes (e.g., poverty, limited English proficiency). As a derived demand, district demand for 12

15 teachers also reflects citizen demand for education. An exogenous set of demand variables (D), such as income, state aid, voter tax share, and variables related to voter preference for education, can be substituted for S in the labor demand equation. Solving the demand equation for W and substituting it into the supply equation results in a reduced form model of teacher qualifications (Q): Q = f (P L, P K, C, A, R, N, Z, D). (1) We assume for simplicity that equation (1) has a constant elasticity functional form, which implies Cobb-Douglas technology, and constant elasticity functions for labor supply and education demand. The reduced form approach limits our ability to identify the structural parameters, such as the effect of recruitment practices on labor supply. The coefficient on R (c R ) in equation (1) is equal to: c R = to (β K /(α L + β K ))α R, (2) where β K is the output elasticity for capital as a percentage of returns to scale, α L is the supply elasticity for teacher qualifications with respect to relative wages, and α R is the effect of recruitment practices on supply of teacher qualifications. While we cannot identify α R, it is likely to be larger than c R ; as long as β K and α L are positive as expected, then (β K /(α L + β K )) will be less than one Factors Affecting District Use of Recruitment Practices Teacher recruitment regressors have to be treated as endogenous variables, because recruitment practices are likely to be chosen when districts make decisions about staffing and budgets. To identify possible instruments, we evaluated superintendents responses to an openended question asking them to list the major constraints faced by your district in recruiting new teachers. Among exogenous factors mentioned by superintendents, rural or remote location was 13

16 the most frequently cited reason. Other limitations mentioned include a small pool of candidates in certain specialties, and resource limitations both in terms of staff time, and budget constraints. Based on these responses and our evaluation of the descriptive results, we hypothesize that district adoption of a recruitment practice is affected by three exogenous factors: enrollment size, county population, and district wealth. First, the most consistent finding in the descriptive tables is that use of practices is strongly related to district enrollment, with adoption of most practices going up with enrollment size. 16 Second, districts with low fiscal capacity (property wealth) may be less willing to invest scarce resources in a broad range of recruitment practices with uncertain benefits; thus, use of most practices is hypothesized to go up with property wealth. Third, districts may have less need to use a broad array of recruitment practices if they are located in a large local labor market with a significant number of potential teachers. We might expect then that use of recruitment practices would decline with the population of the county in which the district is located Measuring Level of Use of Recruitment Practices Since most districts use a bundle of recruitment practices, it is difficult to isolate the impact of any one practice. However, the results in Table 6 suggest that districts can be roughly categorized as heavy, moderate or light users of recruitment practices. For our empirical analysis we develop several classifications for level of use of recruitment practices by districts. First, we categorize districts as broad users of recruitment practices if they use at least one practice from each of the following categories: internet, supply, incentives, and active college strategies. Approximately 52% of districts fall into this category (Table 10). A smaller share of districts (9%) are classified as innovators because they use several non-traditional strategies including at least one strategy from each the following categories: non-traditional recruiting incentives, active 14

17 strategies with non-local colleges, recruitment of alternatively certified teachers, and other internet strategies besides posting job notices on the school website. We label districts as traditional users (19% of districts) if they do not use any non-traditional strategies. The terms nontraditional or innovator are used purely for descriptive purposes to describe practices not used by most school districts, not to signify effectiveness. Another way to classify users is by the number of strategies that they employ. We label districts as heavy users (21% of districts) if they use at least two recruitment strategies from each of the following categories: internet, incentives, and local supply, as well as at least three college strategies. By contrast, districts are classified as light users if they use two college strategies or less, and one strategy or less from each of the other recruitment categories (8% of districts) Data and Measures The independent variables used in the empirical model are generally based on information published by the New York State Education Department (SED). To match the time period used in constructing the teacher qualification factor score, most variables are expressed as a three-year average from 2002 to 2004 (unless noted otherwise). Data from SED includes enrollment counts, the share of elementary school students receiving free lunch, the share of students who are nonwhite, pupil density (pupils per square mile), data on teacher qualifications, adjusted gross income, and regional classifications (downstate, rural, and upstate suburbs). 18 Information on county-level unemployment rates and private salaries (average payroll for professional and technical services) are from the New York State Department of Labor. Market values for property are estimated by the New York Office of Real Property Services. Crime rates are supplied by the New York State Division of Criminal Justice Services. County population is estimated by the U.S. 15

18 Bureau of the Census, and the percentages of the population age 65 years or older, and 5 to 17 years of age are from the 2000 Census of Population. Several variables in the teacher quality model have been dropped because of high collinearity. 19 Table 8 reports descriptive statistics for model variables. 20 <Table 8 about here> 4.4 Empirical Results Models are estimated with linear 2SLS with instruments based on the three factors associated with district adoption of recruitment practices discussed in section 4.2 (enrollment, county population, and per pupil property values). Overidentification tests and weak instrument tests have been run to assess the appropriateness of the instruments. 21 Hypothesis tests are based on robust standard errors. While these estimates should be viewed as exploratory, they may be suggestive of the effect of recruitment practices on teacher qualifications. The full regression results are reported in Table 9, when the measure for broad users of recruitment practices is included in the model. The coefficient on this variable is statistically significant from zero at the 10% level. Because the teacher qualification measure has a standard deviation of 1, the coefficient on the broad users indicates that use of a broad range of recruitment practices is associated with a 0.67 standard deviation increase in teacher qualifications. Coefficients for other variables in the model generally have the expected sign, and several are statistically significant from zero. For example, teacher qualifications are estimated to be positively related to private sector salaries and the county unemployment rate, and negatively related with the share of free lunch students. <Table 9 about here> 16

19 Coefficients on the other recruitment measures are reported in Table 10. The coefficient on the indicator variable for innovative districts (model 2) is 0.90, although it is not statistically significant from zero at conventional levels. Traditional users (districts not using non-traditional practices) are estimated to have teacher qualifications 0.64 standard deviations below districts using at least some of these practices (model 3). A similar story emerges when we look at measures of the number of strategies used. Heavy users of recruitment strategies (model 4) are associated with teacher qualifications that are 0.74 standard deviations above non-heavy users, while lights users (Model 5) have qualifications 1.06 standard deviations below non-light users. 5. Conclusions The survey results have provided a rich picture of the teacher recruiting process in New York State. Most New York school districts advertise in local newspapers, work with local colleges to recruit, post job notices on their school website, recruit substitute teachers, and use extra compensation for extra-curricular or administrative functions as a recruiting incentive. Relatively few districts, on the other hand, advertise outside their local area, work with non-local colleges, search for job candidates on the internet, or offer signing bonuses, assistance with home purchase, or compensation for hard-to-staff fields and schools as recruiting incentives. The most consistent finding from our descriptive analysis is that the use of recruiting incentives goes up with the enrollment size of the school district (except for local newspaper advertising). With regard to fiscal health, high need urban districts are more likely to use some active and innovative strategies than other districts, such as the internet for recruiting, supervisng of student teachers from non-local colleges, and non-traditional recruiting incentives. 17

20 While all districts employ a portfolio of recruitment practices, some districts appear to have a much more diversified portfolio than others. The use of recruitment practices in one category (e.g., college strategies) is positively related to recruitment strategies in other categories (e.g., incentives), particularly for non-traditional recruiting strategies. In other words, some districts are heavy users and some are light users of recruitment practices. The overlap in use of practices makes evaluating the effects of individual practices difficult, but we can categorize districts by the level of use of recruitment practices to determine if there are benefits to utilizing a broad array of recruiting strategies. In the second part of the paper we carry out an analysis of the relationship between level of use of recruitment practices and teacher qualifications. To measure teacher qualifications we develop a composite measure, which includes certification test score performance, selectivity of the college(s) the teacher attended, and certification status in their assignments. The crosssectional nature of the survey does not permit the use of time series data to identify effects of recruitment practices. Instead, we estimate a reduced-form model of teacher qualifications with the recruitment measures treated as endogenous variables. Instruments are selected among exogenous variables associated with the adoption of these practices. While the regression results should be viewed as an exploratory, they provide some of the first evidence available that district recruitment practices affect district teacher qualifications. One consistent finding from the regression analysis is that using a limited set of recruitment practices is negatively related to teacher qualifications. This finding holds for both light users of recruitment practices and for districts not using any non-traditional recruitment practices. The size of the effects appears to be quite large, at least one standard deviation. We also found that broad use of recruitment practices was associated with higher teacher 18

21 qualifications. Focusing on heavy users of practices, particularly non-traditional practices, the results are less definitive. While the coefficients on the measures of innovative users or heavy users are positive, they are not statistically significant at conventional levels. What are the potential implications of these findings for state education policymakers? First, if our findings are correct then state governments may want to target assistance to light (or traditional) users of recruitment practices, primarily small rural districts. If states want to encourage the use of a broader set of recruitment practices by rural districts, they need to help lower the costs of recruitment practices for these districts by providing technical assistance, access to teacher recruitment websites, and financial assistance with recruiting incentives. Regional education organizations, such as New York s BOCES, could assist rural districts to design, manage, and implement recruitment programs. Second, state governments can help to reduce the uncertainty surrounding the benefits of adopting particular recruitment practices by improving the quality of and access to information, including funding program evaluation research to help identify which methods are successful and which are not for particular types of districts; and supporting regional organizations to provide training and assistance in adopting successful strategies. 19

22 References Ballou, Dale Do public schools hire the best applicants? Quarterly Journal of Economics 111 (1): Ballou, Dale, and Michael Podgursky Teacher recruitment and retention in public and private schools. Journal of Policy Analysis and Management 17 (3): Baugh, William, and Joe Stone Mobility and wage equilibrium in the educator labor market. Economics of Education Review 2 (3): Bound, John, David Jaeger, and Regina Baker Problems with instrumental variables estimation when the correlation between the instruments and the endogenous explanatory variables is weak. Journal of the American Statistical Association 90 (June): Boyd, Donald, Hampton Lankford, Susanna Loeb, and James Wyckoff The draw of home: How teachers preferences for proximity disadvantage urban schools. Journal of Policy Analysis and Management 24 (1): Boyd, Donald, Pamela Grossman, Hampton Lankford, Susanna Loeb, and James Wyckoff How changes in entry requirements alter the teacher workforce and affect student achievement. Education Finance and Policy 1 (2): Chamber, Jay Public School Teacher Cost Differences Across the United States: An Analysis to Develop a Teacher Cost Index (TCI). Washington DC: National Center for Education Statistics. Clewell, Beatriz, Katherine Darke, Thonoa Davis-Googe, Laurie Forcier, and Sarah Manes Literature Review on Teacher Recruitment Programs. Washington DC: U.S. Department of Education. Dillman, Donald Mail and internet surveys, The tailored design method. New York: John Wiley and Sons, Inc. Dolton, Peter, and Wilbert van der Klaauw The turnover of teachers: A competing risks explanation. Review of Economics and Statistics 81 (3): Ehrenberg, Ronald, and Dominic Brewer Do school and teacher characteristics matter? Evidence from high school and beyond. Economics of Education Review 13 (1): Falch, Torberg and Bjarne Strom Teacher turnover and non-pecuniary factors. Economics of Education Review 24 (6):

23 Ferguson, Ronald Can schools narrow the black-white test score gap. In The Black-White Test Score Gap ed. Christopher Jencks and Michael Phillips, Washington, DC: The Brookings Institution. Ferguson, Ronald, and Helen Ladd How and why money matters: An analysis of Alabama schools. In Holding Schools Accountable, ed. Helen Ladd, Washington DC: Brookings Institution Press. Glazerman, Steven, Daniel Mayer, and Paul Decker Alternative Routes to teaching: The impacts of Teach for America on student achievement and other outcomes. Journal of Policy Analysis and Management 25 (1): Goldhaber, Daniel The mystery of good teaching. Education Next 2 (1): Teacher licensure tests and student achievement: Is teacher testing an effective policy? Unpublished paper, University of Washington. Goldhaber, Daniel, and Emily Anthony. forthcoming. Can teacher quality be effectively measured? National Board Certification as a signal of effective teaching.review of Economics & Statistics. Goldhaber, Daniel, and Dominic Brewer Does teacher certification matter? High school teacher certification status and student achievement. Educational Evaluation and Policy Analysis 22: Gorsuch, Richard Factor Analysis. London: Lawrence Erlbaum Associates. Greenwald, Robert, Richard Laine and Larry Hedges A meta-analysis of studies of the effects of differential school inputs on student outcomes. Educational Researcher 23 (3): Hanushek, Eric The economics of schooling: Production and efficiency in public schools. Journal of Economic Literature 24 (3): Hanushek, Eric, John Kain, and Steven Rivkin Why public schools lose teachers. NBER Working Paper Series no. 8599, National Bureau of Economic Research, Cambridge, MA. Hanushek, Eric, John Kain, and Steven Rivkin Teachers, schools, and academic achievement. NBER Working Paper Series no. 6691, National Bureau of Economic Research, Cambridge, MA. Hirsch, Eric Teacher Recruitment: Staffing Classrooms with Quality Teachers. Denver, CO: National Conference of State Legislatures. Hussar, William Predicting the need for newly hired teachers in the United States to Washington, DC: National Center for Education Statistics. 21

24 Imazeki, Jennifer Teacher salaries and teacher attrition. Economics of Education Review 24 (4): Kearney, C. Philip National Board Certification: An underutilized resource for New York State? Proceedings from the Symposium on the Teaching Workforce. Albany, NY: Education Finance Research Consortium. Loeb, Susanna How teachers choices affect what a dollar can buy: Wages and quality in K-12 schooling. Proceedings from the Symposium on the Teaching Workforce. Albany, NY: Education Finance Research Consortium. Murnane, Richard, and Randall Olsen The effects of salaries and opportunity costs on length of stay in teaching: Evidence from North Carolina. Journal of Human Resources 25 (1): Murray, Michael The bad, the weak and the ugly: Avoiding the pitfalls of instrumental variable estimation. Unpublished paper, Bates College. Odden, Alan, and Carolyn Kelley Addressing teacher quality and supply through compensation policy. Proceedings from the Symposium on the Teaching Workforce. Albany, NY: Education Finance Research Consortium. Sanders, William, Arnold Saxton, and Sandra Horn The Tennessee value-added assessment system: A quantitative, outcome-based approach to educational measurement. In Grading Teachers, Grading Schools: Is Student Achievement a Valid Evaluation Measure? ed. Jason Millman, Thousand Oaks, CA: Corwin Press. Scafidi, Benjamin, David Sjoquist, and Todd Stinebrickner. forthcoming. Race, poverty, and teacher mobility. Economics of Education Review. Stinebrickner, Todd An empirical investigation of teacher attrition. Economics of Education Review 17 (2): Stone, Joe Value-added assessment: An accountability revolution. In Better Teachers, Better Schools, ed. Marci Kanstoroomm, and Chester Finn, Washington, DC: Thomas B. Fordham Foundation. Strauss, Robert, Lori Bowes, Mindy Marks, and Mark Plesko Teacher Performance and Selection in Pennsylvania. Report prepared for the Pennsylvania State Board of Education Improving teacher preparation and selection: lessons from the Pennsylvania experience. Economics of Education Review 19 (4):

25 Table 1 Advertising for Teachers in Media Outlets by District Characteristics District Enrollment Low Medium High All Districts N Date of first advertisements (months) Date of typical offer (months) Percent of responses indicating "most" or "all" advertising placed in these media outlets: Local newspapers/periodicals (within 50 miles) Other advertising: Other newspapers/periodicals in New York Newspapers/periodicals in other states Education trade publications/periodicals Radio/television SED Need/Resource Capacity Categories High Need Urban High Need Rural Average Need Low Need Date of first advertisements (months) Date of typical offer (months) Percent of responses indicating "most" or "all" advertising placed in these media outlets: Local newspapers/periodicals (within 50 miles) Other advertising: Other newspapers/periodicals in New York Newspapers/periodicals in other states Education trade publications/periodicals Radio/television Note: "Low" is below 25th percentile (approximately 1,000 students), "medium" is 25th to 75th percentile, and "high" is above the 75th percentile (approximately 3,500 students). Months are defined as 1=January, 2=February, etc. Bold and italics indicate a statistically significant difference among district enrollment size categories or among need/resource capacity categories (10% level) using a chi-square test. 23

26 Table 2 Location of Colleges Where Districts Using Recruitment Strategies by District Characteristics Local Colleges Non-Local Colleges District Enrollment Low Medium High All Districts Low Medium High All Districts All strategies List job openings: Post job notices at the college Advertise in placement newsletter distributed by college Active strategies: Visit campus to actively recruit job candidates Contact specific college faculty to identify potential job candidates Supervise student teachers from the college High Need Urban High Need Rural High Need Urban High Need Rural SED Need/Resource Capacity Categories Average Need Low Need Average Need Low Need All strategies List job openings: Post job notices at the college Advertise in placement newsletter distributed by college Active strategies: Visit campus to actively recruit job candidates Contact specific college faculty to identify potential job candidates Supervise student teachers from the college Note: Sample size is 494. Percent of responses using a particular recruitment strategy with a college. Local colleges are defined as colleges within 50 miles. "Low" is below 25th percentile (approximately 1,000 students), "medium" is 25th to 75th percentile, and "high" is above the 75th percentile (approximately 3,500 students). Bold and italics indicate a statistically significant difference among district enrollment size categories or among need/resource capacity categories (10% level) using a chi-square test. 24

27 Table 3 Use of the Internet for Recruitment by District Characteristics District Enrollment Low (Using BOCES) Low Medium High All Districts Total number of internet strategies used Percent of responses: Uses internet to recruit teachers Posts job openings on the internet: School district website Online recruitment websites targeted to teachers General online recruitment websites Other uses of the internet for recruitment: Searches for candidates on a recruitment website Allows candidates to submit applications online SED Need/Resource Capacity Categories High Need Urban High Need Rural Average Need Low Need Total number of internet strategies used Percent of responses: Uses internet to recruit teachers Posts job openings on the internet: School district website Online recruitment websites targeted to teachers General online recruitment websites Other uses of the internet for recruitment: Searches for candidates on a recruitment website Allows candidates to submit applications online Note: Sample size is 494. Percent of responses indicating use of internet to recruit teachers. "Low" is below 25th percentile (approximately 1,000 students), "medium" is 25th to 75th percentile, and "high" is above the 75th percentile (approximately 3,500 students). Sum of internet strategies is a simple count of the number of strategies (0 to 5). Bold and italics indicate a statistically significant difference among district enrollment size categories or among need/resource capacity categories (10% level) using a chi-square test. 25